Research And Improvement Of Granular Association Rules Algorithm

Information overload in the Internet age has gradually affected all aspects of people's daily lives.Recommendation system solves the problem of information overload and is therefore widely used.Recommendation system can provide personalized and differentiated services,but there are still a series of challenges,such as cold start.Granular association rules algorithm can solve the cold start problem of recommendation system.Granular association rules algorithm is based on granular computing and rough sets,and uses four measures to mine the hidden patterns in the multi-relational data table.The rules mined by granular association rules algorithm are richer in semantics than traditional association rules.Granular association rules algorithm solves the problem of difficult to mine due to multi-valued attributes and numerical data in the mining process by using Scaling-based method and discretization method.Granular association rules recommend corresponding rules for users with different attributes by matching user granules.In addition,according to the granular confidence and significance measures,granular association rules algorithm provides Top-k recommendation rules for different users to improve the effectiveness of the recommendation system.Granular association rule algorithm has many advantages,but it is based on Apriori algorithm when constructing rules,so it is difficult to avoid the problem of rule redundancy caused by Aprioir algorithm.This paper introduces the concept of maximal frequent itemsets based on the granular association rules to improve this problem.The experimental results show that the improved algorithm reduces the number of rules and runs faster than the original algorithm.Although the improved algorithm has fewer rules than the original algorithm,there is a small amount of rule loss in the improved algorithm.In order to evaluate the impact of rule loss on the whole recommendation system,the accuracy of the algorithm before and after the improvement is compared under different thresholds and different division ratios of training sets and testing sets.The experimental results show that the improved algorithm has higher accuracy than the original algorithm.The rules used by granular association rules in Top-k recommendation are based on binary relationship.Therefore,the probability of users adopting rules in recommendation is only considered,and whether users like them or not is not considered.In order to solve the problem caused by binary relationship,This paper considers the granular association rule Top-k recommendation algorithm from the perspective of data cubes.The experimental results show that the improved algorithm has higher recommendation significance and accuracy than the original granular association rule algorithm under different thresholds and different k values.In summary,this paper proposes a corresponding solution to the problem of rule redundancy in the granular association rule algorithm and the lack of interest in the Top-k recommendation.The experimental results verify the effectiveness of the improved algorithm.